Exploring spatial dependency and heterogeneity in forest land dynamics in the randstad metropolitan region: A combined spatial nonlinear modeling approach

Achieving sustainable urban development requires careful management of forested areas in rapidly changing metropolitan regions. However, this task is limited by the insufficient understanding of the spatial and nonlinear associations between forest land changes, and socioeconomic and biophysical factors. To address this challenge, in this study, we applied a combined spatial nonlinear approach, bringing together a global model (Gaussian-process tree-boosting, GPBoost) with a local model (geographically-weighted random forest, GWRF). This approach provides both a region-wide perspective, revealing spatial dependency and identifying priority areas for forest land management, and a local perspective, revealing spatial heterogeneity and specific local factors that influence forest land changes. This study was conducted in the context of the Randstad metropolitan region in the Netherlands, a densely urbanized area experiencing intense land-use conflicts. The results reveal that combining global and local perspectives is crucial to fully capture spatial associations across scales and prevent spatial mismatches between demand and supply for forest land management. Moreover, the combined approach highlights the importance of avoiding oversimplified linear assumptions when developing forest land protection policies.